Haskins Laboratories Status Report on Speech Research1991, SR-105/106, 9-22

The Role of Working Memory in Reading Disability*

Susan Amanda Bradyt

Other presentations have focussed on the impor­tance of phonological awareness in reading acqui­sition and on the central role Isabelle Libermanhas played in theoretical and empirical advancesin this area. It was noted that she led the way inrecognizing the cognitive demands of reading: thatreading, in contrast to speaking and listening, re­quires explicit awareness of phonological seg­ments and that this awareness is difficult toachieve given the embedded nature of phonemesin syllables. In this paper I will show that Dr.Liberman also led the way in investigating howmetaphonological abilities relate to underlyingphonological processes. She was among the first toidentify the need to understand the organizationand functioning of the language system in order toexplain sources of difficulty for poor readers.Together with Donald Shankweiler and severalstudents, Isabelle Liberman conducted insightfuland elegant research on the working memorydeficits of poor readers, discovering that phonolog­ical processes are implicated here, as well as inthe deficits in phonological awareness.

In my presentation I would like to accomplishthree things: First, I will take note of the largebody of evidence that deficits in a specifically ver­bal form of working memory are associated withreading problems, and to point out Dr. Liberman'senormous contribution to our understanding of thesource of these limitations in language processing.

I gratefully acknowledge the helpful comments of AnneFowler and Donald Shankweiler on an earlier draft. Myresearch was supported in part by a Program Project Grant(HD-01994) to Haskins Laboratories from the NationalInstitute of Child health and Human Development. Some ofthe material presented in this paper was adapted from thearticle, Reading ability and short-term memory: The role ofphonological processing, by Michele Rapala and Susan Brady,1990, in Reading and Writing, (2), pp. 1-25. Dordrecht, TheNetherlands: Kluwer Academic Publishers. Copyright 1990 byKluwer Academic Publishers. Adapted by permission.

9

Second, I will consider evidence (both old andnew) that the efficiency of phonological processesis an important limiting factor in working memorycapacity and that poor readers often haveinefficient phonological processes. Third, I willdiscuss current evidence pertaining to possiblecausal links between the phonological processesunderlying phonological awareness, verbal work­ing memory, and lexical access.

The association of verbal workingmemory deficits with reading disabilityWhile deficiency in metaphonological awareness

is certainly the .language factor most stronglyimplicated in reading disability, there is reason tobelieve that difficulties in awareness betokenmore basic problems in language use. At the levelof underlying language processes, perhaps themost striking characteristic of poor readers is thecommon occurrence of verbal memory problems.Teachers often comment on the difficulties poorreaders have retaining information for even briefperiods of time, and there now exists a massivenumber of research studies reporting anassociation between reading disability and short­term memory limitations. When given a short listof digits, or letters, or words, or even nameablepictures to recall, poor readers recall fewer itemsthan do good readers. The generality of therelationship between reading difficulty andworking memory is highlighted by the finding thatit holds both for readers of alphabetic writingsystems and for people in Eastern Asia learningsyllabic and logographic scripts (Mann, 1985; Ren& Mattingly, 1990). Evidence also emerges fromthe unusual condition of hyperIexia: Healy, Aram,and Horowitz (1982) presented findings on 12hyperlexic children who despite very low cognitivefunctioning in almost every area were gooddecoders. Performance on working memory tasksstood out as one of the few cognitive strengths ofthese children.

10 Brady

Evidence for a causal role of memory in readingperformance comes from a small number of pre­diction studies that found memory capacity inkindergarten to be significantly related to latersuccess at learning to read. For example, Share,Jorm, Maclean, and Matthews (1984) reported ona longitudinal study of 543 children and found acorrelation of about .4 between kindergarten per­formance on sentence memory tasks and readinglevel at the end of first grade. In a subsequentanalysis of these children, Jorm, Share, Maclean,and Matthews (1986) subdivided the children whobecame poor readers (now followed to the end ofthird grade) into those with a specific readingproblem and those who had reading problems andalso had low IQs. For both groups languagedeficits, including poor memory performance, were 'apparent at the kindergarten phase prior toreading acquisition. The low IQ children, in addi­tion to their language impairments, had otherproblems such as poor performance on perceptual­motor tasks and on a nonverbal auditory task.

The consequences of a reduced capacity inworking memory may be extensive. Some havesuggested that limited capacity may make it moredifficult to discover and master metaphonologicalskills. Evidence compatible with this is presentedby Wagner, Balthazor, Hurley, Morgan, Rashotte,Shaner, Simmons, and Stage (1987). In this studywith 111 kindergarten nonreaders, the resultsbest fit a model specifying a single latent factor forphonological awareness and for verbal workingmemory. Others, such as Perfetti and Lesgold-(1977; 1979), have noted that deficits in memorymay make it more difficult to learn how to decode.A recent study by Dreyer, 1989, reports a correla­tion of .65 between performance on a recall taskand decoding skill. A further impact of limitedmemory resources on reading has been suggestedby Perfetti (1985) who hypothesized that for theslow, beginning decoder limits in working memorymay be used up getting to the words of the text.Once the words have been decoded, insufficientresources are left for other higher language pro­cesses. This is consistent with the observationthat children may fail to comprehend a sentencein text even when they manage to decode all thewords it contains. Similarly, Liberman and hercolleagues have stressed the integrative functionof working memory and considered that its chiefcontribution would be to facilitate syntactic andother sentence-level processes (Liberman,Shankweiler, Liberman, Fowler, & Fischer, 1977;Mann, Liberman, & Shankweiler, 1980;Shankweiler, Liberman, Mark, Fowler, & Fischer,

1979). In keeping with this, Mann, Shankweiler,and Smith (1984), Smith, Macaruso, Shankweiler,and Crain (1989), and Fowler (1988) have foundevidence that poor reader's difficulties withsentence level tasks stem in part from deficits inworking memory.

Although memory difficulties in poor readers arecommonplace, they are not universal. In an impor­tant study by Torgesen and Houck (1980), twogroups of learning disabled children wereidentified, only one of which had memory impair­ments. Further, the proportion of variance be­tween good and poor readers accounted for byworking memory performance is generallymarkedly lower (approximately 10%) than theproportion accounted for by phonological aware­ness abilitin3 (approximately 40-70%). Wagner(1988) reported on a meta-analysis of longitudinalcorrelational studies and training studies thatfinds relatively independent causal roles formetaphonological processes and for working mem­ory. Thus the exact role of working memorydeficits in reading disability remains unclear. Onthe one hand, the occurrence of these difficultiesin poor readers is widespread, having been foundacross language and writing systems, in diversepopulations, and to be present in those destined tobe poor readers prior to reading acquisition. Onthe other hand, it must be acknowledged thatmemory limitations are not as consistently or asstrongly associated with reading disability as aredeficits in metaphonological skills.

These facts warrant a searching examination ofworking memory processes and their role inreading. Ultimately, we will need to understandthe interplay between metaphonological processesand memory processes to fully comprehend therequirements of reading acquisition and the basesof reading failure.

A group of investigators from HaskinsLaboratories began this enterprise by exploringthe basis and extent of memory deficits for poorreaders (Liberman et aI., 1977; Shankweiler et aI.,1979). Drawing on the evidence that informationin verbal working memory is retained in aphonological or speech code, they hypothesizedthat poor readers have a specific difficulty withthe use of phonological representations in workingmemory, and went on to demonstrate that goodreaders, like adults (Baddeley, 1966; Conrad,1972), find lists of nonrhyming words easier torecall correctly than rhyming lists. Presumably,because in the nonrhyming case the internalphonological or speech representations for thewords are very distinct, whereas for the rhyming

The Role of Working Memory in Reading Disability 11

words they are so similar to each other that it iseasier for them to be confused. Unlike the goodreaders, poor readers didn't show this pattern.They were not adversely affected by rhyme;performance on rhyming lists was not appreciablydifferent from performance on nonrhyming lists.These results were interpreted to suggest thatpoor readers are not as adept as good readers atforming phonological representations.

In subsequent work the Haskins researchersand others have found the rhyme effect in goodand poor readers to be affected by both subjectand task factors (Brady, Mann, & Schmidt, 1987;Ellis, 1980; Hall, Wilson, Humphreys, Tinzmann,& Bowyer, 1983; Watkins, Watkins, & Crowder,1974). Nonetheless, the line of research on therhyme effect, or phonological similarity effect,paved the way in pointing to the role ofphonological processes in the memory and readingdifficulties of poor readers.

Using this paradigm, Shankweiler et al. (1979),made a further noteworthy observation that thememory differences between good readers andpoor readers were apparent whether visualpresentation was used, or auditory presentation.They suggested that poor readers have a generalproblem with the use of a phonological code,independent of how the material is presented, nota difficulty that is restricted to the reading processitself. In contrast, they, and others (Katz,Shankweiler, & Liberman, 1981; Liberman,Mann, Shankweiler, & Werfelman, 1982;Vellutino, Pruzek, Steger, & Meshoulam, 1973),found that poor readers do not perform less wellthan good readers on short-term memory taskswith nonspeech stimuli such as doodle drawingsor photographs of strangers. With these stimuli,not easily given a phonological label, no significantdifferences in performance between good and poorreaders were observed. These findings supportedthe conclusion that poor readers do not suffer froma general memory impairment. Rather they areoften deficient in the ability to rememberlinguistic material, however it is presented, andthe problem appears to be related to phonologicalprocesses involved in encoding or storing verbalinformation.

Explanations of developmentaland reading-group differences in

memory spanAt this point it wasn't apparent whether poor

readers were employing some other codingstrategy to retain items in STM, such as a visualor semantic strategy, or whether they were using

phonological codes, but doing so inefficiently. Longbefore, Conrad (1971) had suggested that childrenyounger than six did not use phonological codingin memory. However, more recent research byAlegria and Pignot (1979) found the rhyme effectto be present in children as young as four, and,indeed, phonological coding in working memorywould seem to be essential for a child to learn tospeak a language. Yet, for some reason, poorreaders show the rhyme effect at a later age thando children who are better readers. For example,using a cross-sectional design, Olson, Davidson,Kliegl, and Davies (1984) reported that byadoles­cence poor readers demonstrated the phonologicalsimilarity effect, though they still showed lowerlevels of recall than good readers (see alsoJohnston, 1982, and Siegel & Linder, 1984).

These findings suggest that factors whichcontribute to developmental differences in verbalSTM performance may also account for readinggroup differences in linguistic memory. I shouldnote that Dempster (1981), reviewing numerousstudies, estimated that span approximately triplesfrom an average of slightly more than two items atage two to an average of nearly seven items atadulthood. Children who are poor readerssystematically lag behind their age mates in recalllevel. Three kinds of explanations have beenoffered to account for developmental and readinggroup differences in verbal STM performance. Iwould like to describe all three, but I'll be briefabout the first two, which seem inadequate, andI'll elaborate more on the third position; anoperational efficiency view that emphasizes therole ofphonological processes.

Mnemonic strategies. First, the differential useof mnemonic strategies has been offered as anexplanation for developmental increments in STMspan. According to this view, children becomeincreasingly aware that certain strategies willenhance recall, and are thus increasingly able toemploy an appropriate technique. It is clear thatmnemonic strategy use does advance as childrenbecome older. These advances include morefrequent spontaneous use of rehearsal strategies,such as the way we repeat a phone number toremember it briefly, and better use of imposed orsubjective organization.

Good and poor reader differences have also beenexplained in this way. Tarver, Hallahan,Kauffman, & Ball, 1976, and Torgesen, 1977,reported that good readers are more likely to useself-initiated rehearsal than children with readingproblems. And Torgesen (1978-79) reported thatgood readers are more likely to use a chunking

12 Brady

strategy or to consciously impose an organi­zational plan on materials to be recalled. Butadditional findings, as well as interpretivedifficulties, suggest that differences in mnemonicstrategy use is not the sole basis for individualdifferences in memory function. Lange (1978)confirmed that adults and adolescents were morelikely than younger children to make use oforganizational principles, but differences in thisstrategy were not discerned in children from 5 to12 years of age. Yet this is a period during whichshort-term memory span has been found toincrease dramatically (Dempster, 1981).

A second conflicting piece of evidence for the roleofmnemonic strategies in developmental increasesin span comes from a series of experiments whichhave equated children's use of c(!rtain controlprocesses during STM tasks. For example, Samuel(1978) grouped items in internal patterns in recalllists. Children ranging in age from six to nineteenall benefitted equally from that groupingtechnique, so the age differences were maintained(see also Huttenlocher & Burke, 1976, and Engle& Marshall, 1983).

Similar results have been obtained when thesubjects were good readers and poor readers. Thatis, when reading groups are equated on the use ofstrategies, differences in recall span still occur.Thus while the use of mnemonic strategiesimproves as individuals get older, it does notappear to be the central factor either indevelopmental or in reading-group differences.

Capacity. A second approach to developmentalchanges in memory span has postulated an actualdifference in capacity of short-term memory.Proponents of capacity explanations hold that acertain number of memory slots are availableprior to the presentation of any stimuli.Developmental differences in span are attributedto the presence of a greater number of these slotsas a function of age: so a two-year-old might have2 slots, and adults might have 7 slots (Pascual­Leone, 1970, and Halford & Wilson, 1980). Thisposition is consistent with the observation ofdevelopmental increases in recall, and also fits theobserved differences in STM span among good andpoor readers.

However, a series of experiments conducted byBaddeley, Thomson, and Buchanan (1975) demon­strated that the number of items that can berecalled is not fixed for a given individual. Thenumber varies with the length of the stimulus (innumber of syllables). Thus, temporal duration ofthe item list proved to be a strong determinant ofspan. Similarly, Ellis and Hennelley (1980) did

experiments with bilingual Welsh students andshowed that these subjects had a longer digit spanwhen the digits were given in English than inWelsh, maintaining that the differences wereattributable to the fact that Welsh numbers take alonger time to say. In addition, as will bediscussed below, increases in memory span withage can be offset ifunfamiliar stimuli are used forrecall. Thus it seems than an explanation in termsof a fixed number of slots cannot account forimportant individual differences in STMperformance.

Operational efficiency. The third position I wishto present, which I find a more promisingexplanation, is an operational efficiency view thatproposes that memory is served by a limited ca­pacity system and that memory operations such asencoding and retrieval become more efficient withrelevant experience. As a result, the amount ofoperating resources needed to complete short-termmemory tasks decreases with age and there is afunctional increase in storage capacity. I find ituseful to view the working memory system as apie: ifperception or encoding requires one quarterof the pie, three-quarters will be left over for re­call. If one gets better at encoding, more resourceswill be available for memory. My version of thishypothesis is that the difficulty observed in encod­ing, phonological information is not restricted tomemory tasks but occurs at a more abstract level,whenever it is necessary to create and maintain aphonological representation. To evaluate thishypothesis one can look for parallels between per­formance on tasks involving speech perception,speech production, and verbal short-term memory.I will begin with some observations from the de­velopmentalliterature and I will review here somestudies with single age groups that seem relevant.

First, with adult subjects it has been observedthat if the encoding requirements on a memorytask are made more difficult by changing theperceptual demands, then short-term recallsuffers. For example, Rabbitt (1968) had adultslistening to digits in white noise. In one conditionthey were asked to repeat individual items, inanother condition they had to recall lists of digits.Noise levels that caused no effect on perception ofthe individual items, significantly impaired recallof the same stimuli. Recall of the lists not in noisewas greater than recall of the lists with noise.Adding noise and increasing the perceptualdifficulty adversely effected memory. Likewise,Luce, Feustel and Pisoni (1983) demonstrated thatadults have poorer recall of lists of syntheticallyspoken items than of lists of natural speech, and

The Role of Working Memory in Reading Disability 13

Mattingly, Studdert-Kennedy, and Magen (1983)reported poorer recall if the words were spoken indifferent dialects. These three studies illustratedifferent ways of making the encoding demandsmore difficult, but all are compatible with the viewthat when more resources are used up forencoding, less are available for storage or recall,hence there is a trade off.

A second line of evidence for the relevance ofunderlying phonological processes to workingmemory functioning comes from a report by Lockeand Scott (1979) that children with articulationdisorders also have impaired short-term memoryperformance. Similarly, Ellis (1979) reviewsfindings of concordant error patterns by adults inspontaneous speech production and in short-termmemory tasks.

The third source of evidence come from indica­tions that for both adults and children, theintrinsic efficiency of phonological processescorrelates positively with memory span. Thussignificant correlations are obtained between howfast one can speak and how much is recalled on ashort-term memory task. This has been foundboth in adult studies (Baddeley, et at, 1975;Hoosian, 1982; Naveh-Benjamin & Ayres, 1986;Nicolson, 1981) and in developmental studies.Hulme, Thomson, Muir, and Lawrence (1984)tested individuals from four years of age toadulthood and obtained a linear relationshipbetween maximal speaking rate and recallperformance. A particularly compelling set offindings comes from a study by Case, Kurland,and Goldberg in 1982. Testing three-to-six yearold children, all of whom are unlikely to usemnemonic strategies, these authors observed astrong correlation between how rapidly thechildren could repeat the test words and the sizeof their memory span. The older children, whocould articulate faster, recalled more of the words.In a convincing test of this relationship, Case etat found that when adults' speaking rate wasslowed to the rate for six-year-old children bygiving the adults more difficult items, memoryspan with these stimuli also dropped to the six­year-old level. Although the encoding processesinvolved in memory were only evaluated by word­repetition speed, these results indicate that devel­opmental increases in memory span may be linkedto the efficiency of related phonological processes.

Let us consider the reading-group literature toexamine whether the verbal memory differences ofgood and poor readers might also be accounted forby the efficiency of underlying phonologicalprocesses. This remains to be fully addressed:

speech perception, verbal memory, and speechproduction abilities in good and poor readers havegenerally been looked at in isolation, and sorelations among these processes have not beenexamined in depth. I'll discuss current findingsaddressing the operational efficiency hypothesisand will then present findings from a study withMichele Rapala (Rapala & Brady, 1990).

First, I have noted investigations which found acorrespondence in poor readers between less effec­tive use of phonological coding in short-termmemory (indicated by a lack or reduction of thephonological similarity effect) and reduced mem­ory span (Mann et al., 1980; Mark, Shankweiler,Liberman, & Fowler, 1977; Olson et at, 1984;Shankweiler, et al., 1979; Siegel & Linder, 1984).Related to this, error analyses have indicated thatwhile both reading groups use phonological codingstrategies, poor readers are less accurate. In 1983Donald Shankweiler, Virginia Mann and I firstnoted this, looking at the errors made by third­grade good and poor readers on lists of non­rhyming words (Brady, Shankweiler, & Mann,1983). The response sequences included items, forboth good readers and poor readers, that had notoccurred in the original strings. In the vast major­ity of cases, these errors were recombinations ofphonological components that were present in theinitial sequence (for example, for the target itemstrain and plate, several subjects reported "trait"and "plane"). These transposition errors occurredfor both reading groups, but more frequently forthe poor readers. Thus, the poor readers wereclearly using a phonological coding strategy, buthaving more difficulty than the good readers inretaining the correct combination of phonologicalsegments. In a follow-up study, Brady et at (1987)used lists of nonsense syllables to assess the inci­dence and circumstances of such errors withgreater precision. We constructed lists in whichthe items shared zero, one, or two phonologicalfeatures. Across lists of syllables, the order of thestimuli also varied systematically so we could askwhether adjacent items were more likely to betransposed than were nonadjacent ones, andwhether the effect of shared features is greatestfor immediately adjacent segments. As expected,poor readers made more errors than good readers,but for both groups transposition errors accountedfor the majority of errors. For both reading groups,these more often involved swaps from adjacentsyllables and there was a significant effect ofphonological feature similarity, with a greaternumber of transpositions occurring between sylla­bles that had features in common. (The effects of

14 Brady

feature similarity and of adjacency seemed to beindependent.) These kinds of errors suggest thatthe inferior performance of poor readers is not theconsequence of a different coding strategy, but re­lates to differences in the formation or storage ofphonological representations. Thus, we again havea correspondence between less effective coding inSTM and poorer recall.

A second area of investigation has demonstratedspeech perception deficits for poor readers,possibly reflecting a general difficulty in encodinglanguage. Different paradigms have been usedsuch as word repetition tasks and categoricalperception tasks, but the common finding is thatpoor readers are less accurate (Apthorp, 1988;Brady et al., 1983; Brady, Poggie, & Rapala, 1989;Godfrey, Syrdal-Lasky, Millay, & Knox, 1981;Goetzinger, Dirks, & Baer, 1960; Palley, 1986;Read, personal communication; Snowling, 1981;Snowling, Goulandris, Bowlby, & Howell, 1986;Werker & Tees, 1987.

The task has to be somewhat demanding in or­der to detect lower performance by poor readers.For example, Brady et al. (1983) found poorreaders to be worse at identifying monosyllabicwords in noise, but not monosyllabic wordspresented without noise. In subsequent studies weand other have also found poor readers to do lesswell on clearly presented words if the phonologicaldemands are increased either by lengthening thestimuli by using multisyllabic words or bydecreasing item familiarity by using pseudowords.The evidence with multisyllabic real words, aswell as earlier findings that word frequency ofstimuli has a comparable effect on both readinggroups (Brady et al., 1983), argues against anexplanation that poor readers' difficulties inphonological processing are confined to nonwords(cf., Snowling et al., 1986). It is unclear, however,whether the problem repeating words arisesduring the perception or production components ofsuch tasks, or to the common requirement offormulating a phonological representation.Indeed, with respect to speech production, thereare reports in the clinical literature thatindividuals with reading difficulty often displaymisarticulations in their speech (Blalock, 1982;Chasty, 1986; Johnson & Mylebust, 1967; Klein,1986), as well as empirical demonstrations thatdyslexics are slower and less accurate at repeatingphrases (Catts, 1986; Catts, submitted: though forthis task memory requirements may be a factor.)Finally, additional evidence of a link betweenphonological processes and memory capabilitieswith respect to reading level has been tentatively

supported by positive correlations between howfast children can name lists of digits and theirmemory span (Spring & Capps, 1974; Spring &Perry, 1983; Torgesen & Houck, 1980.) Thenaming task involves speech production but alsorequires retrieving the names of the digits. Thus,more than one factor may be contributing to thecorrelation with memory.

In our own work we've found poor readers to beas fast as good readers at single word repetition,but not as accurate (Brady et al., 1989; Rapala &Brady, 1990). A recent study by Stanovich,Nathan, and Zolman (1988) using a reading-agematch design makes ,the point that speech of ar­ticulation is strongly age-dependent. Childrenmatched on reading ability who were in the third,fifth, and seventh grades differed markedly inspeed of repetition of simple words though theirmemory spans were comparable. This would seemto undercut the importance of the link betweenreading ability and articulation, and by extension,the role of speed of formulating phonological rep­resentations in memory span. Yet the studies withadults cited earlier that report a continuing asso­ciation between speed of articulating, memoryspan and reading ability raise doubts that devel­opmental factors artifactually account for the cor­respondence between memory span and articula­tion found by Hulme et al. (1984). Likewise, theexperimental design of Case et al. (1982) also pre­sents convincing evidence of a significant relationbetween encoding processes and memory span. Insum, phonological difficulties in memory, inspeech perception, and in speech production havebeen observed in children with reading difficultiesand there have been occasional reports of correla­tions between these measures. Relations amongthe underlying phonological processes need to bemore closely investigated, both in developmentalstudies and in reading-group comparisons. In ad­dition, we need to disentangle the factors of speedand accuracy of articulation as they relate tomemory span and reading ability.

To review, in seeking an explanation of theshort-term memory deficits characteristic of poorreaders, I have considered the explanations pro­posed for developmental changes in memory per­formance. The explanation in terms of mnemonicstrategy use did not seem adequate. The use ofmnemonic strategies does increase developmen­tally and is sometimes observed to be superior ingood readers as compared to poor readers, butnoteworthy changes in memory span occur duringyears in which changes in mnemonic strategiesare not evident. Further, when different age

The Role of Working Memory in Reading Disability 15

groups or reading groups are induced to use thesame strategy, differences in span are still pre­sent. The capacity view, maintaining that the ac­tual number of slots in memory increases, is con­tradicted by evidence that the length and familiar­ity of items are important determinants of recall.The third position maintains that verbal workingmemory is a limited capacity system and that theefficiency of encoding and retrieving limit the re­sources available for retaining information. Boththe developmental research and the reading-groupresearch point toward a role of the efficiency ofphonological processing in memory capacity.However, in each field the questions have been in­completely addressed, and some of the findingsappear to be in conflict. The developmental stud­ies have focussed on the correspondence betweenspeaking rate and memory, emphasizing the im­portance of speed of articulation. The readingstudies have more extensively investigatedphonological processes in speech perception andspeech production that may relate to memoryperformance, but have generally examined eacharea in isolation.

A study by Michele Rapala and me (Rapala &Brady, 1990) was designed to investigate morethoroughly whether developmental and readinggroup differences in verbal STM can be accountedfor by differences in the efficiency of relatedphonological processes. Using an extensive batteryof tests, a cross-sectional developmental study wascarried out with 4 1/2 year olds, 6 1/2 year oldsand 8 1/2 year olds to examine the associationbetween verbal STM and phonological processes inspeech perception and speech production.Complementing the developmental study, acomparison of8 1/2 year old good readers and poorreaders was conducted using the same test batteryto assess directly whether poor readers' difficultiesin STM are associated with deficits in otherphonological processes.

The measures included a verbal short-termmemory task using lists of words and a non-verbalmemory task known as the Corsi Blocks test(Corsi, 1972). In that task, nine identical blackblocks are scattered on a platform and the exper­imenter points to a number of blocks, in turn. Thesubject then must reproduce this sequential pat­tern. We also included a number of phonologicaltasks. There were two speech repetition tasks;repetition of monosyllabic and multisyllabicwords. The children were told to say each wordcarefully but as quickly as possible. To control forthe role of perception in the first task, we alsoincluded tasks in which the child would hear a

tone and was asked to respond with a particularword (in one condition the monosyllable "cat"; inanother condition the multisyllable "banana").Here no speech signal has to be encoded, so weeliminated the speech perception requirement. Inthe word repetition and control tasks, responseswere scored for speed of onset and for accuracy.Lastly, we had a production measure in which thechild was told to repeat six times in succession atwo-syllable tongue twister such as "sishi" or"bublu." On this task, the time to produce the sixrepetitions was measured and the accuracy ofeach syllable was recorded.

We obtained significant correlations between allthe measures in our developmental study, but ofcourse everything improves with age. To checkthat the variable of age, or some general cognitivefactor that improves with age, wasn't the actualbasis for the significant relations, we conducted aseries of correlational analyses with the effects ofage controlled. When age is controlled (see Table1) a significant relationship continues to existbetween VSTM and the other phonological pro­cesses. The negative direction of the coefficientsindicates that as the time to produce stimuli de­creases, and as the number of errors decreases,verbal short-term memory span increases. Thecontrol measures which tested speed of articula­tion without a perceptual component did not corre­late significantly after age was partially out.Perhaps these tasks weren't sufficiently sensitive;alternatively, the processes tapped by these mea­sures are less closely related to memory.

Ifwe look at the results between these measuresand non-verbal memory (see Table 2), again withage partialled out, the overall patterns is verydifferent. One variable produced a significantcorrelation, but with this many comparisons thatcould be a chance result. The lack of a generalrelationship between nonverbal memory andphonological processes contrasts sharply with theconsistent association found between verbal recalland the other phonological measures. Thisindicates that while there are age-relatedimprovements in a variety of cognitive skills, theobserved relationship between verbal STM andthe other phonological measures expressly reflectsshared linguistic processing factors. Indeed, theresults of a discriminant function analysis,conducted with the nine language measuresindicated that one significant discriminantfunction accounted for 83.4 percent of the totalvariance. This result suggests the tests of speechperception, speech production, and verbal memorytap a single underlying dimension.

16 Brady

Table 1. Developmental study: First order correlations (age partialled) between verbal short-term memory and eachofeight phonological variables (N=74).

Variable

Accuracy measures

Speech Perception: Errors on Monosyllabic WordsSpeech Perception: Errors On Multisyllabic WordsTongue Twister Errors

Speed measures

Speech Perception: RT on Monosyllabic WordsSpeech Perception: RT on Multisyllabic WordsTongue Twister SpeedControl: Monosyllabic RTControl: Multisyllabic

*p< .05

Partialr

-.25*-.43*-.42*

-.25*-32*-35*-.21-.17

Table 2. Developmental study: First order correlations (age partialled) between verbal short-term memory span andnine language measures (N=74).

Variable

VerbalSTM

Accuracy Measures

Speech Perception: Errors on Monosyllabic WordsSpeech Perception: Errors On Multisyllabic WordsTongue Twister Errors

Speed measures

Speech Perception: Reaction Time (RT) on Monosyllabic WordsSpeech Perception: RT on Multisyllabic WordsTongue Twister RTControl: Monosyllabic RTControl: Multisyllabic

*p< .05

Partial r

.20

-.09-.12-.07

-.24*-.18-.08-.01-.02

On comparing the 8 1/2 year old good and poorreaders, we found, as have others, that the groupsdiffered on verbal memory but not on nonverbalmemory. We also found that memory performancecorrelated significantly with the error scores formultisyllabic words and for tongue twisters, andthat a substantial proportion of the variance inVSTM span could be accounted for by the accuracyof phonological processing (VSTM, multisyllabic

errors: r2 =.25; VSTM, tongue twister errors: r2 =.15).

In sum, it is likely that verbal working memoryplays a major role in reading, whether memorydeficits, per se, emerge as a primary factor or as acontributing factor of reading disability.Therefore, it is essential to understand the func­tioning of the phonological system serving verbalworking memory. The present findings are com-

The Role of Working Menwry in Reading Disability 17

patible with the view that: a) poor readers aredeficient in phonological processing; b) that phono­logical limitations lead, in tum, to limitations inthe efficient use of working memory; and, c) thatnonverbal memory processes are served by sepa­rate cognitive functions. However, this is a veryminimal conceptual framework, and much workremains to be done on the nature of the workingmemory system and on the factors accounting fordifferences in capacity. For example, as noted ear­lier, the role of both speed and accuracy in phono­logical processing needs to be further studied todetermine how these processing variables relate tothe functioning of the phonological system and tothe construct of efficiency. In addition, it isimportant to study further the basis and extent ofthe commonalities across speech perception,speech production and verbal memory. Ourfindings fit the hypothesis that the need to encodethe stimulus phonologically is shared across allthe tasks, and that this may be the basis of thecorrelations we obtained. Yet we appreciate thatother factors such as rehearsal and retrieval needto be be studied in this light to evaluatealternative explanations.

In addition, the present findings suggest itwould also be worthwhile to explore furtherwhether the deficits of poor readers on verbalmemory tasks are to be explained in terms oflower efficiency on a continuum of normalphonological processing. Lastly, it will beimportant to relate the deficits of poor readers onthese more basic phonological processes to therobust evidence that poor readers also havedeficits in metaphonological awareness. In thenext section we tum to this question, reviewingcurrent findings on the relations among variousphonological processes.

What are the relations amongphonological processes?

In other chapters in Brady and Shankweiler (inpress), we have focussed on phonologicalawareness and on verbal working memory aspotential factors in reading disability. Before weconsider how these may be related, I must remindyou that a third area of language function, lexicalaccess, is often also associated with readingability. Naming, as a test of lexical access,requires the subject to retrieve a phonologicallabel in response to visual stimuli such as colors,numbers, letters, or pictured objects. Poor readerstend to be slower on tasks requiring the rapidnaming of visual stimuli and they also have been

reported to make more errors in retrievingphonologically complex labels (i.e., words such asthermometer or stethoscope (Catts, 1986; Katz,1985). The strength of the association betweenreading ability and measures of lexical access hasbeen found to depend on the particular measurestaken and other factors such as word frequency,age, and task variables. The correspondence isstrongest when tested in kindergarten, and whenthe task is to scan and label an array of highfrequency items selected from a finite pool (e.g.,Wolf, 1986). With older children, reading-groupdifferences are consistently found only when theitems to be labeled are orthographic symbols (e.g.,Blachman, 1983; Katz & Shankweiler, 1985; Wolfet at, 1986). (See Stanovich, 1985, and Wagner &Torgesen, 1987, for reviews.)

Although each of these abilities draws on phono­logical representation, we are not yet in a positionto explicate the nature of the underlying relationsamong phonological awareness, verbal workingmemory, and lexical access. Research addressingthis question has been limited, piecemeal, andcontradictory. If one considers how the phonologi­cal processes may be related, there are five logicalpossibilities, as Wagner and his colleagues indi­cate (Wagner et at, 1987): 1) There may be aunitary deficit common to metaphonological pro­cesses, phonological coding in working memory,and phonological coding in lexical access; 2)Though all are phonological, each may be a sepa­rate ability, with three independent processes im­plicated; 3) The metaphonological processes maybe independent of the other two. Wagner et at,1987, term this pattern of organization "anawareness versus use" classification. Theremaining two combinations represent the otherpossible permutations of two latent abilities; 4)Phonological awareness and verbal working mem­ory may be tied to a common factor. In this view,phonological awareness tasks rely heavily onthe efficiency of coding in working memory, butuse of phonological information to access storeditems in the lexicon taps separate processes; 5)Phonological awareness and lexical access may belinked, perhaps by some relation between aware­ness of phonological structure and organization orutilization of phonological information in the lexi­con. Phonological coding to maintain informationin working memory would then constitute aseparate aspect of processing.

The unitary deficit view is attractive in light ofthe compelling evidence that poor readers havedeficits largely confined to the phonologicaldomain of language. It is appealing to propose

18 Brady

that the three classes of phonological abilitieshave some underlying common factor thataccounts for the individual deficits. All draw onthe same underlying knowledge base: thephonological component of the languageapparatus. In this way, there would be uniquerequirements for metaphonological, workingmemory, and lexical processes, but the underlyingphonological representation might be inadequateand yield difficulties in each of the other areas(Liberman & Shankweiler, 1985).

So far, the studies that have attempted toexamine abilities in all three phonological areashave not supported a unitary deficit, but there hasnot otherwise been much agreement betweenthem. One study, a meta-analysis of the results oflongitudinal correlational studies and trainingstudies, indicated relatively independent causalroles in the acquisition of reading skills for thethree phonological abilities we have identified(Wagner, 1988). The second study, conducted with51 adults (including familial dyslexics, clinicidentified dyslexics, and normally reading adults),also supported a multiple factors model(Pennington, Van Orden, Kirson, & Haith, inpress), though individual tasks loaded somewhatdifferently on the three factors. The third, a studyof preschoolers' phonological processing abilities(Wager et aI., 1987) reported that at thisprereading stage two coherent phonologicalabilities were evident: 1) memory and awarenesshaving a common factor and 2) lexical processes asa separate component.

Other investigations have examined therelations between only two of the threephonological abilities; here too the results havebeen inconsistent. One line of research hasattempted to examine whether the deficits inphonological awareness and memory relate to acommon factor. In one view, phonologicalawareness tasks may rely heavily on the efficiencyof coding in working memory. And indeed, thoseawareness measures that most strongly relate toreading ability (Stanovich, Cunningham &Cramer, 1984; Yopp, 1988) generally involvecomparing a number of stimuli, manipulatingsegments, or deleting segments. All of these tasksrequire greater working memory involvementthan less discriminating phonological awarenessmeasures such as rhyme generation. Conversely,the development of phonological awareness mayfacilitate the ability to use phonological segmentsin working memory, as opposed to largerarticulatory units such as syllables (see Fowler,this volume). Thus, one research goal has been to

investigate whether there is a common deficit andto contrast metacognitive processes from morebasic language processes. It is known, forinstance, that in the absence of readinginstruction, abilities to isolate and manipulatephonological segments are limited (Morais, Cary,Alegria, & Bertelson, 1979; Read, Zhang, Nie, &Ding, 1986). And yet it is presumed that theunderlying phonological processes involved inordinary speaking and listening have developednormally. Further, in studies of adults who failedto develop skill in reading, phonological awarenessdeficits are present, while difficulties in the otherphonological abilities are not as apparent as inyoung poor readers (see Pennington et aI., inpress). These age-specific differences among poorreaders lead to the hypothesis that perhapsmetaphonological processes are distinct from theunderlying phonological domain and that theassociations between awareness and memory inyoung children merely reflect simultaneousincreases in development. An altemative may bethat the emergence of phonological awareness andsubsequent reading acquisition depend on acertain minimum of working memory efficiency,beyond which individual differences in memorycapacity are less crucial. In fact, research inchildren designed to evaluate the correspondencehas yielded mixed results. Thus, Mann andLiberman (1984), Fowler (1988) and Goldstein(1976) reported significant correlations betweenawareness and memory tasks. On the other hand,several studies failed to find significantcorrelations between awareness and memorymeasures (Alegria, Pignot, & Morais, 1982;Blachman, 1983; Mann, 1984), or evidence (fromfactor analysis) that they load on a single factor(Mann & Ditunno, in press).

Nevertheless, as I've argued in the precedingsection, a link appears to be justified among themore basic language tasks associated with readingdisability. Verbal memory, phonologicalperception, speech production, and lexical accessall require the creation of a phonologicalrepresentation, whether this representation isinitially generated internally or activated byincoming stimuli. As noted, we and others haveobtained significant correlations between theaccuracy of phonological processes in perceptionand production and the capacity of workingmemory; this relationship obtains develop­mentally as well as in comparisons of good readersand poor readers. In contrast, these processes arenot related to nonverbal memory performance (cf.Rapala & Brady, 1990). Significant correlations

The Role of Working Memory in Retuiing Disability 19

have also been reported between naming speedand memory span (Spring & Capps, 1974; Spring& Perry, 1983; Torgesen & Houck, 1980), althoughStanovich (1985) notes inconsistencies in thisoutcome. Along similar lines, Elbro (1988) findsthat poor performance in discriminating minimalpairs of words in noise is associated with slowlexical assess. Thus while we have evidence ofcorrespondence between the underlying processes,these mayor may not be separable fromphonological awareness.

In sum, it seems premature to attempt to statehow the various phonological processes are relatedto one another. At present, the field ischaracterized by lack of accord. Theoretical andempirical evidence can be gathered for a numberof the potential patterns of relations among thephonological processes. Gains will no doubt bemade, as they were in previous decades (e.g.,VeIlutino, 1979), by improving how we ask theimportant questions. Several methodologicalshortcomings are now recognized as detractingfrom the merit of many studies. For example, thelack of convergent results may stem from thefrequent use of a single task for each construct,which may not provide an adequately valid,reliable, or sensitive measure. Thus in futurework it will be important to include multiplemeasures of each construct to attain more robustinformation. In addition, it will be important tofactor out general cognitive ability since thisaffects performance on almost all cognitivemeasures, and may inflate the associationbetween any of the phonological processing tasks.Third, many of the studies have not had asufficient number of subjects to warrant thestatistical procedures used and to obtain reliableinformation on the questions addressed. Inaddition, we need conceptual advances so that weknow better how to evaluate the processingrequirements ofvarious tasks. At present, in eacharea of phonological processing several measuresare used and, for the most part, we know littleabout their comparability. For example, memorytasks may involve a continuous presentation taskin which only the last few of a numerous andindeterminate number of stimuli are requested forrecall, they may entail recall of entire strings, orof individual items. The stimuli may be digits,words, nonwords, or pictures. The taskrequirements are not constant, and performanceon them may not vary uniformly with respect toreading ability.

Selection criteria for subjects also remains athorny issue. We still have the concerns about how

to assess reading ability and IQ, and aboutwhether we should assess performance in otherareas of achievement (e.g., math) or of cognitivefunctions (e.g., attention). The more recent vogueof reading-age matches only exacerbates theseconcerns. It has been very profitable to comparethe particular reading skills of individuals whoare at an equivalent reading level in order toaddress questions about deviant reading patterns,but I am concerned that when we extend the useof this paradigm to study underlying cognitiveprocesses (e.g., verbal memory) that the dualdiversity of age and cognitive ability makes itdifficult to match subjects appropriately or tointerpret the outcome (see Shankweiler, Crain,Macaruso, & Brady, in press, for furtherdiscussion).

All of these points bring me to the somewhattiring conclusion that in the reading field we needto refine our knowledge of the individualconstructs of speech perception, speechproduction, verbal working memory, and lexicalaccess. In other chapeters in the Brady andShankweiler (in press) volume, we can see thatthe effort to fine-tune our understanding of theconstruct of phonological awareness is alreadyunderway. I don't want to imply that we shouldn'tat the same time attempt to test hypotheses abouthow the whole picture fits together, but it will beimportant to be cautious about interpretingindividual studies. As Stanovich (1985) discusses,we need to adhere to the principle of convergingevidence. The power of this principle has beenawesome concerning the evidence that readingproblems are associated with language deficitsand, in particular, with phonological awareness.When we ask, as we have been here, how thedifferent phonological processes relate,convergence is lacking and we need to ask why.

In closing, in this paper I have attempted to ac­complish three goals. In the first section, thewidespread association of verbal short-termmemory deficits with reading disability was brieflyreviewed. The prevalence of this associationunderscores the need to examine the factors thatcontribute to working memory performance. Inthe second section of the paper I argued thatstudies of memory development and of memorydeficits in poor readers provide tantalizingindications that efficiency of phonological codingmay be an important factor in memoryperformance. In the third section I raised thequestion of how working memory processes arerelated to other phonological processes,phonological awareness and lexical access that

20 Brady

have also been associated with reading deficits. Atpresent, attempts to delineate the connectionsbetween them have yielded contradictoryindications. Conceptual and methodological ad­vances, which hinge on the continued cooperationof disciplines involved in the study of language,will be needed to advance our grasp of theseissues.

REFERENCESAlegria, J., Pignot, E. (1979). Genetic aspects of verbal mediation

in memory. Child Development, 50,235-238.. Alegria, J., Pignot, E., &: Morais, J. (1982). Phonetic analysis of

speech and memory codes in beginning readers. Memory iiiCogniticm, 10 451-456.

Apthorp, H. (1988). P1umetic coding in retUling disllbled amtmunityaJllege stwlents. Unpublished doctoral dissertation, Universityof Connecticut, Storrs.

Baddeley, A. D. (1966). Short-term memory for word sequences asa function of acoustic, semantic and formal similarity. QuarterlyJournal ofExperimenUil Psychology, 18, 362-365.

Baddeley, A, Thomson, N., &: Buchanan, M. (1975). Word lengthand the structure of short-term memory. Jou17Ul1 of VerbalLearning iii Verbal BdultJior, 14, 575-589.

Blachman, B. (1983). Are we assessing the linguistic factors criticalin early reading? Annals of Dyslexi/l,33, 91-Hl9.

Blalock, J. (1982). Persistent auditory language deficits in adultswith learning disability. Journal of Learning Disabilities, 15,604­fJ.YJ.

Brady, S., Mann, V., &: Schmidt, R. (1987). Errors in short-termmemory for good and poor readers. Memory iii Cognition, 15,444-453.

Brady, S., POggie, E., &: Rapala, M. M. (1990). Speech repetitionabilities in children who differ in reading skill. LanglollZge andSpeech, 32(2), 109-122

Brady, S. A., Shankweiler, D., &: Mann, V. A. (1983). Speechperception and memory coding in relation to reading ability.Journal of ExperimenUil Child Psychology, 35, 345-367.

Case, R., Kurland, D. M., &: Goldberg, J. (1982). Operationalefficiency and the growth of short-term memory span. Journal ofE:rperimenUil Child Psychology, 33, 386-404.

Catts, H. (submitted). Phonological deficits in reading disorderchildren.

Catts, H. (1986). Speech production/phonological deficits inreading-disordered children. Journal of Learning Disabilities, 19,504-508.

Chasty, H. (1986). What is dyslexia? A developmental languageperspective. In M. Snowling (Ed.), Children's written IImg&l4gediffictdties (pp. 11-27). Windsor, Berkshire, UK: Nfer-Nelson.

Conrad, R. (1971). The chronology of the development of covertspeech in children. DeuelopmenUil Learning and Vemu BehaTJior,5,398-405.

Corsi, P. M. (1972). Human memory and the medial temporal region ofthe brain. Unpublished doctoral dissertation, McGill University.

Dempster, F. N. (1981). Memory span: Sources of individual anddevelopmental differences. Psychologiad Bulletin, 89, 63-100.

Dreyer, L. (1989). 1M relaticmship of children's phonologiad memoryto decoding and reading ability. Unpublished doctoraldissertation, Columbia University.

Elbro, C. (1988). Morphemic awareness among dyslexics. Paperpresented at the Seventh International Symposium onDevelopmental Dyslexia and Dysphasia. Academia RodinensisPro Remediatone. Stockholm: Wenner-Gren Center.

Ellis, A W. 1979). Speech production and shot-term memory. In J.Morton &: J. c., Marshall (Eds.), Psycholinguistics 2: Structure andprocess. Cambridge, MA: MIT Press.

Ellis, A. W. (1980). Errors in speech and short-term memory: Theefforts of phonetic similarity &: syllable position. Journal of .vemzz Learning iii Verbal Behavior, 19, 624-634.

Ellis, N. c., &: Hennelly, R. A (1980). A bilingual word-lengtheffect: Implications for intelligence testing and the relative easeof mental calculation in Welsh and English. British Jou17Ul1 ofPsychology 71, 43-52.

Engle, R. W., &: Marshall, K. (1983). Do developmental changes indigit span result from acquisition strategies? Journal ofExperimental Child Psychology, 3,429-436.

Fowler, A. (in press). How early phonological development mightset the stage for phoneme awareness. In S. Brady &: D. P.Shankweiler (Eds.), Phonologiad processes in literacy: A tribute toIsabelle Y. Liberman. Hillsdale, NJ: Lawrence ErlbaumAssociates.

Fowler, A. (1988). Grammaticality judgments and reading skill ingrade 2. Anllllis of Dyslexia, 38,73-84.

Godfrey, J. J., Syrdal-Lasky, A. K., Millay, K. K., &: Knox, C. C.(1981). Performance of dyslexic children on speech perceptiontests. Journal of ExperimenUil Child Psychology, 32, 401-424.

Goldstein, D. M. (1976). Cognitive-linguistic functioning andlearning to read in preschoolers. Journal of ExperimentalPsychology, 68,~.

Goetzinger, C., Dirks, D., &: Baer, C. J. (1960). Auditorydiscrimination and visual perception in good and poor readers.AM4lls ofOtology, Rhinology, and Laryngology, 69, 121-136.

Halford, G. S., &c Wilson, W. H. (1980). A category theoryapproach to cognitive development. Cognitiue Psychology, 12,3!»411.

Hall, J. W., Wilson, K. P., Humphreys, M. S., Tinzmann, M. B., &:Bowyer, P. M. (1983). Phonemic-similarity effects in good vs.poor readers. Memory iii Cognition, 51, 520-527.

Healy, J., Aram, D., Horowitz, S., &: Kessler, J. (1982). A study ofhyperlexia. BNin and Lang&l4ge, 17,1-23.

Hoosian, R. (1982). Correlation between pronunciation speed anddigit span size. Perceptual and Motor Skills, 55,1128.

Hulme, c., Thomson, N., Muir, c., &: Lawrence, A (1984). Speechrate and the development of short-term memory span. Journalof ExperimenUil Child Psychology, 38, 241-253.

Huttenlocher, R., &c Burke, D. (1976). Why does memory spanincrease with age? COGPSY, 8, 1-31.

Johnston, R. (1982). Phonological coding in dyslexic readers.British Journal ofPsychology, 73, 455-460.

Johnson, D. Myklebust, H. (1967). Learning disabilities: Educaticmprinciples and practices. New York: Grone and Stratton.

Jorm, A. F., Share, D. L., Maclean, R. &: Matthews, R. (1986).Cognitive factors at school entry predictive of specific readingretardation and general reading backwardness: A researchnote. Journal ofChild Psychology and Psychiatry, ZI, 45-55.

Katz, R. B. &: Shankweiler, D. (1985). Phonological deficiencies inchildren with reading disability: Evidence from an objectnaming task. Cognition, 22, 225-257.

Katz, R., Shankweiler, D. &: Uberman, I. (1981). Memory for itemorder and phonetic recoding in the beginning reader. Journal ofExperimerIUIl Child Psychology, 32, 474-484.

Klein, H. (1986). The assessment and management of somepersisting difficulties in learning disabled children. In M.Snowling (Ed), Children's written lang&l4ge difficulties (pp. 59-79).Windsor, Berkshire, UK: Nfer-Nelson.

Lange, G. (1978). Organization-related processes in children'srecall. In P. A. Ornstein (Ed.), Memory tleTJelopment in children.Hillsdale, NJ: Lawrence Erlbaum Associates.

The Role of Working Memory in Reading Disability 21

Liberman, I. Y., Mann, V. A, Shankweiler, D., & Werfelman, M.(1982). Children's memory for recurring linguistic andnonlinguistic material in relation to reading ability. Cortex,18,367-375.

Liberman, I. Y., & Shankweiler, D. (1982). Phonology and theproblems of learning to read and write. Remedial Imd SpecialEduauion, 6, 8-17.

Liberman, I. Y., Shankweiler, D., Liberman, A. M., Fowler, C. A,& Fischer, F. W. (1977). Phonetic segmentation and recoding inthe beginning reader. In A S. Reber & D. 1.. Scarborough (Eds.),TOWilrd il psychology of retUling: The proaedings of tM CUNYConferences (pp. 207-225). Hillsdale, NJ: Lawrence ErlbaumAssociates.

Locke, J. L., & Scott, K; K. (1979). Phonetically mediated recall inthe phonetically disordered child. Joumlll of CommuniClltionDisorders, 12, 125-131.

Luce, P. A Feustel, T. C., & Pisani, D. B. (1983). Capacitydemands in short-term memory for synthetic and naturalspeech. HUmiln FilCtors, 25,17-32.

Mann, V. A (1984). Longitudinal prediction and prevention ofreading difficulty. Annllls of Dysle:cill, 34,117-137.

Mann, V. A (1985). A cross-language perspective in the relationbetween temporary memory skills and early reading ability.Remedial ilnd Special EdUC/ltion, 6, 37-42.

Mann, V. A & Ditunno, P. (1990). Phonological deficiencies:Effective predictors of future reading problems. In G. Pavlikes(Ed.), Dyslexill: A neuropS1jclwlogiall ilnd laming perspectilN!. NewYork: Wiley.

Mann, V. A, & Liberman, I. Y. (1984). Phonological awarenessand verbal short-term memory: Can they presage early readingproblems? Journlll of Letaming Disllbilit~s, 17, 592-599.

Mann, V. A, Liberman, I. Y., & Shankweiler, D. (1980). Children'smemory for sentences and wordstrings in relation to readingability. Memory & Cognition, 8, 329-335.

Mann, V. A, Shankweiler, D., & Smith, S. T. (1984). Theassociation between comprehension of spoken sentences andearly reading ability: The role of phonetic representation.Journlll ofChild LAn8Uilge, 11,627-643.

Mark, L. S., Shankweiler, D., Liberman, I. Y., & Fowler, C. A(1977). Phonetic recoding and reading difficulty in beginningreaders. Memory & Cognition, 5, 623-629.

Mattingly, I. G., Studdert-Kennedy, M. & Magen, H. (1983, May).Paper presented at the meeting of the Acoustical Society ofAmerica, Cincinnati, Ohio.

Morais, J., Cary, L., Alegria, J., & Bertelson, P. (1979). Doesawareness of speech as a sequence of phonemes arisespontaneously? Cognition, 7, 323-331.

Narveh-Benjamin, M. & Ayres, T. (1986). Digit span, reading rate,and linguistic relativity. QUilrterly JOUTnilI of ExperimentillPsychology, 38A, 739-751.

Nicolson, R. (1981). The relationship between memory span andprocessing speed. In M. Friedman, J. P. Das, & N. O'Connor(Eds.), Intelligence ilnd lellrning (pp. 179-184). Plenum Press.

Olson, R. K., Davidson, B. J., Kliegl, R., & Davies, S. E. (1984).Development of phonetic memory in disabled and normalreaders, Journlll of E:rperinrentlll PS1jclwlogy, 37, 187-206.

Pascual-Leone, J. (1970). A mathematical model for the transitionrule in Piaget's developmental stages. Actil PsychologiCil, 32, 301­345.

Palley, S. (1986). Speech perception in dysle:cic children Unpublisheddoctoral dissertation, The Qty University of New York.

Pennington, B. F., Van Orden, G., Kirson, D., & Haith, M. (inpress). What is the causal relation between verbal STM prob­lems and dyslexia? In S. Brady & D. P. Shankweiler (Eds.),PlwnologiClll processes in literilC1j: A tribute to I~lle Y. Libermlzn.Hillsdale, NJ: Lawrence Erlbaum Associates.

Perfetti, C. A (1985). Resuling lIbility. New York: Oxford UniversityPress.

Perfetti, C. A., & Lesgold, A M. (1979). Coding and comprehen­sion in skilled reading and implications for reading instruction.In L. B. Resnick & P. A Weaver (Eds.), Theory ilnd prilCtice ofttlr/y retUli1lg. Hillsdale, NJ: Lawrence Erlbaum Associates.

Perfetti, C. A., & Lesgold, A. M. (1979). Discourse comprehensionand sources of individual differences. Cognitive processes incomprehension. In M. A. Just & P. A Carpenter (Eds.),Cognitit1e processes in comprelunsion (pp. 141-183). Hillsdale, NJ:Lawrence Erlbaum Associates.

Rabbitt, P. M. A. (1968). Channel-capacity, intelligibility andimmediate memory. QUilrterly Joumill of ExperimentillPsychology, 20, 24-248.

Rapala, M. & Brady, S. (in press). Reading ability and short-termmemory: The role of phonological processing. Rellding ilndWriting.

Read, c., Zhang, Y., Nie, H., & Ding, B. (1986). The ability tomanipulate speech sounds depends on knowing alphabetictranscription. Cognition, 24, 31-44.

Ren, N., & Mattingly, I, (1990). Short-term serial recallperformance by good and poor readers of Chinese. HilSlcinsLAlxJTiltoms stiltus Report on Speech Rest!ilrch, sR-103/104, 153­164.

Samuel, A G. (1978). Organization versus retrieval factors in thedevelopment of digit span. JOUTnilI of Experimentill ChildPS1jchology, 26, 308-319.

Shankweiler, D., Crain, S., Brady, S., & Macaruso, P. (1990).Identifying the causes of reading disability. In P. B. Gough(Eds.), Reilding ilcquisition. Hillsdale, NJ: Lawrence ErlbaumAssociates.

Shankweiler, D., Liberman, I. Y., Mark, L. S., Fowler, C. A, &;fischer, F. W. (1979). The speech code and learning to read.Journlll of E:rperimentlll Psychology: HUmiln 1.ellming and Memory,5,531-545.

Share, D., Jorm, A, Maclean, R., & Matthews, R. (1984). Sources ofindividual differences in reading acquisition. Joumill ofEduauional PS1jchology, 76,1309-1324.

Siegel, L S., & Linder, B. A. (1984). Short term memory processesin children with reading and arithmetic learning disabilities.Deuelopmentlll Psychology, 20, 2QO.207.

Smith, S. T., Macaruso, P., Shankweiler, D., & Crain, S. (1989).Syntactic comprehension in young poor readers. AppliedPS1jcholinguistics.

Snowling, M. (1981). Phonemic deficits in developmentaldyslexia. PS1jclwlogiall Rest!ilrch, 43, 219-234.

Snowling, M., Goulandris, N., Bowlby, M., &: Howell, P. (1986).Segmentation and speech perception in relation to readingskill: A developmental analysis. Journlll of E:rperimentlll ChildPsychology, 41, 489-507.

Spring, c., & Capps, C. (1974). Encoding speed, rehearsal andprobed recall of dyslexic boys. Journlll of EduCiltio1Ull Psyclwlogy,66,780-786.

Spring, c., &: Perry, L. (1983). Naming speed and serial recall inpoor and adequate readers. Contemporilry EduciltionillPsyc1wZogy,8,141-145.

Stanovich, K.E. (1985). Explaining the variance in reading abilityin terms of psychological processes: What have we learned?An1Ulls of Dysle:cill, 35, 67-96.

Stanovich, K. E., Cunningham, A. E., & Cramer, B. B. (1984).Assessing phonological awareness in kindergarten children:Issues of task comparability. JouTnilI of Experimentlll ChildPS1jchology,38,175-19O.

Stanovich, K. E., Nathan, R. G., & Zolman, J. E~ (1988). The devel­opmentallag hypotheses in reading: Longitudinal and matchedreading-level comparisons. Child Development, 59, 71-86.

22 Brady

Tarvar, S. G., Hallahan, D. P., Kauffman, ]. M., & Ball, D. W.(1976). Verbal rehearsal and selective attention in children withlearning disabilities. A developmental lag. Joul7l41 ofExperimental Child Psychology, 22, 375-385.

Torgesen, ]. K. (1977). Memorization processes in readingdisabled children. Joul7l4l of Educatimull Psyclwlogy, 69,571-578.

Torgesen, ]. K. (1978-79). Performance of reading disabledchildren on serial memory tasks: A selective review of recentresearch. &uling ReSi!llTch Quarterly, 14, 57~7.

Torgesen,]. K. & Houck, D. G. (1980). Processing deficiencies oflearning-disabled children who perform poorly on the digitspan test. Joul7l4l of EducationlZl Psychology, 72, 141-160.

Vellutino, F. R. (1979). Dyslexia: Theory 4nd research. Cambridge,MA: MIT Press.

Vellutino, F. R., Pruzek, R., Steger, ]., & Meshoulam, U. (1973).Immediate visual recall in poor and normal readers as a func­tion of orthographic-linguistic familiarity. Cortex, 8, 106-118.

Wagner, R. K. (1988). Causal relations between the developmentof phonological processing abilities and the acquisition ofreading skills: A meta-analysis. MerriU-Ptalmer Quarterly, 34,261-279.

Wagner, R., Balthazor, M., Hurley, S., Morgan, S., Rashotte,C.,Shamer, R., Simmons, K. & Stage, S. (1987). The nature ofprereaders' phonological processing abilities. CognitiTJeDeTJtlopment, 2, 355-373.

Wagner, R. K., & Torgesen,]. (1987). The nature of phonologicalprocessing in the acquisition of reading skills. PsychologicalBullttin, 101,192-212

Watkins, M. ]., Watkins, O. c., & Crowder, R. G. (1974). Themodality effect in free and serial recall as a function ofphonological similarity. Journal of Verbtal wl7ling 4nd VerbtalBthIITJior, 13, 430-447.

Werker, ]. & Tees, R. (1987). Speech perception in severelydisabled and average reading children. Canadian Journal ofPsychology, 41, 48-61.

Wolf, M. (1984). Naming, reading and the dyslexias: Alongitudinal overview. Annals of Dyslaia, 34,78-115.

Wolf, M., Bally, H., & Morris, R. (1986). Automaticity, retrievalprocesses, and reading. A longitudinal study in average andimpaired readers. Child DtTJtlopmtnt, 57, 988-1000.

Yopp, H. (1988). The validity and reliability of phonemicawareness tests. &uling ReStilrch Quarterly, 23,159-177.

FOOTNOTES"To appear in S. Brady & D. P. Shankweiler (Eds.), Plwnologiclll

processes in literllcy: A tribute to ISIlbtllt Y. Liberrrll3n. Hillsdale, N]:Lawrence Erlbaum Associates (in press).

tAlso Department of Psychology, University of Rhode Island